Article Text

Download PDFPDF
#228 A comparison between semiquantitative and digital image analysis for the assessment of tumor-infiltrating lymphocytes in high-grade serous ovarian carcinoma
  1. Simona Miceska1,2,
  2. Erik Skof1,
  3. Dineo Khabele2,
  4. Praveen Krishnamoorthy3 and
  5. Veronika Kloboves Prevodnik1
  1. 1Institute of Oncology Ljubljana, Ljubljana, Slovenia
  2. 2Washington University School of Medicine in St. Louis, St. Louis, USA
  3. 3Washington University Center for Cellular Imaging, St. Louis, USA


Introduction/Background The assessment of tumor-infiltrating lymphocytes (TILs) as a prognostic variable in high-grade serous ovarian carcinoma (HGSC) has not seen broad adoption due to a lack of standardization. The International TIL Working Group has standardized semiquantitative TIL scoring, but only for breast cancer. In the era of artificial intelligence, digital imaging image platforms have also been widely used to overcome the challenges of TIL assessment. We aimed to compare TIL scores in HGSC achieved with semiquantitative analysis (SQA) based on breast cancer guidelines and digital image analysis (DIA).

Methodology Formalin-fixed paraffin-embedded tumor sections of patients diagnosed with primary HGSC, immunohistochemically stained with CD3, CD4, and CD8 antibodies, were used. Intra-tumoral and stromal TILs were scored separately. For SQA, breast cancer guidelines were adopted, and for DIA, an algorithm using AI-driven precision pathology software VisioPharm was created. In both assessment methods, the result was calculated as a percentage of positively stained intra-tumoral TIL area per intratumoral area and stromal TIL area per stromal area. Assessments from both methods were compared using Bland Altman, Cronbach’s α, and Spearman’s rank tests.

Results The Bland-Altman analysis showed a percentage difference in TILs assessed with SQA and DIA of 8.4% (HR (95% CI): 1.999, range 0.485-3.505, p=0.010). Moreover, internal consistency between SQA and DIA assessed with Cronbach’s α was very good (α=0.848). A significantly moderate positive monotonic relationship was also observed with Spearman’s rank correlation test (p<0.001; rho=0.581).

Conclusion We observed strong agreement between SQA and DIA methods in analyzing TILs in HGSC, suggesting that the DIA could be a reliable alternative for TIL scoring in comparison to SQA.

Disclosures The authors have no disclosures.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.